AI safety researcher: governments building ASI won't magically solve alignment

davidmanheim · x · 2026-09-24

David Manheim argues we should stop advanced AI training as soon as compliance is verifiable, and that governments building ASI can't magically solve alignment. He notes similar ideas have come from SPAR and RAND researchers. The quoted debate weighs infra-level training restrictions via interconnects: restricting only a few risks power concentration and state misuse, while banning everyone means technological stasis and poor enforceability.

Original post →

More from AGI Musings

AGI Musings channel →